| 15 Feb 2026 |
 Tom | yeah, it seems to be working for me with 1.57.2-dev | 21:41:14 |
| Tom | But looking at tcpdump it doesn't seem like you can even easily tell (if at all) it's a networkmanager dhcp request. So matching based on version seems even more out of the question.
(You could ofc set ipv4.dhcp-vendor-class-identifier but that doesn't seem scalable) | 21:59:19 |
| 16 Feb 2026 |
|  zimward changed their display name from zimward to zim. | 13:49:31 |
| zimward changed their display name from zim to zimward. | 13:50:38 |
|  pneumatic changed their profile picture. | 15:17:47 |
| 17 Feb 2026 |
|  lanestolen set a profile picture. | 15:00:21 |
| lanestolen changed their profile picture. | 15:01:25 |
|  @michaelgrahamevans:matrix.org left the room. | 18:42:42 |
| 18 Feb 2026 |
|  antifuchs joined the room. | 02:00:38 |
| 19 Feb 2026 |
|  ginkogruen joined the room. | 01:30:04 |
|  Lan Tian changed their profile picture. | 08:00:27 |
|  Wilkuu [SNT] {main: @wilkuu:wilkuu.xyz} changed their display name from wilkuu to Wilkuu [SNT] {main: @wilkuu:wilkuu.xyz}. | 21:08:12 |
| 20 Feb 2026 |
 Sandro 🐧 | General KDE Connect breaks :) | 02:36:21 |
| Sandro 🐧 | whose support need some training that NAT64 is not IPv4 NAT :) | 02:36:59 |
| 21 Feb 2026 |
|  h7x4 changed their profile picture. | 19:25:18 |
| h7x4 changed their profile picture. | 19:30:11 |
| h7x4 changed their profile picture. | 19:33:35 |
| 22 Feb 2026 |
| antifuchs | hah, so I have just learned that nixos-nftables-firewall is pretty suboptimal in terms of performance | 01:49:39 |
| antifuchs | a machine that handles 10Gb/s of traffic easily will struggle with this zone config that I applied to it just now (about 170 handles) | 01:51:06 |
|  Haze joined the room. | 02:54:08 |
| antifuchs | (Update, no, that’s not it, the NIC just didn’t have GRO turned on) | 05:07:04 |
|  problems changed their profile picture. | 22:11:02 |
|  mei 🌒& changed their profile picture. | 22:53:36 |
| antifuchs | so, somewhat serious question though, I don't think boot.kernel.sysfs.class.net.eno8np0.device.sriov_numvfs = 32; is working; the path never triggers, even though the file exists. when I manually start the service, it applies the setting; is that boot.kernel.sysfs handling broken somehow, or am I expecting it to do something impossible? | 23:35:29 |
| 23 Feb 2026 |
| Sandro 🐧 | I had before that I needed to manually load a kernel module early on (as in with the option for it) to make the sysfs paths available when the sysfs service starts | 13:02:08 |
| antifuchs | huh, do you remember what kernel module that is? | 16:29:16 |
|  dent set a profile picture. | 22:58:22 |
| 24 Feb 2026 |
|  @cdepillabout:matrix.org left the room. | 07:39:00 |
| 25 Feb 2026 |
|  @renesat:matrix.org left the room. | 00:48:57 |
| Sandro 🐧 | https://gitea.c3d2.de/c3d2/nix-config/src/branch/master/modules/microvm-defaults.nix#L16-L18
initrd.kernelModules = [
# required for net.netfilter.nf_conntrack_max appearing in sysfs early at boot
"nf_conntrack"
];
| 19:42:01 |
